MONITORING OF ENGINE OIL DEGRADATION AND POSSIBLITIES OF LIFE PREDICTIONS IN COMBUSTION ENGINE

The article entitled Monitoring of engine oil degradation and possibilities of life prediction in combustion engine deals with chronological monitoring of engine oil on the monitored object - a passenger car with a petrol engine. The research concerns the basic physico-chemical parameters of motor oil, where it discusses the operational factors that contribute to its degradation. The theoretical part of the thesis deals with the analysis of the current state of the problem in the chemical composition of engine oils, analysis of the current state of contact indicators of oil quality in lubrication systems of internal combustion engines and analysis of contactless systems "live" evaluating engine oil quality during vehicle operation. The research part of the work includes the collection of operational data, laboratory analysis of oil samples and statistical processing of the results of tribodiagnostic monitoring. This article discusses the 1st phase of extensive long-term research in the field of tribology and operation of the Mitsubishi Lancer 1.5 Inform motor vehicle


INTRODUCTION
The main goal of the research is to determine the most vulnerable chemical parameters of motor oil in terms of degradation and their mathematical description with subsequent prediction of service life in motor oil. Another important goal is to determine a suitable predictive model and its implementation in practice passenger car Mitsubishi LANCER 1.5 Inform with gasoline engine 4A91 (80 kW). The vehicle was regularly monitored for almost 3 years -from 01. april 2019 to 30. november 2021. During this period, they covered 32,113 km and 2,027 starts/routes. Each start of the vehicle was consistently recorded in the vehicle's operating log, in which the necessary data were collected. During this interval, three sets of the same Valvoline Syn Power 5W-40 oil fill were tested on the vehicle.

DESIGN AND INSTALLATION OF EXTERNAL MEASURING STATION
Part of the research was long-term monitoring of a real vehicle (Fig. 1), which was the subject of daily use by a physical person.  The means of transport was operated in accordance with road traffic rules for the performance of everyday tasks and was regularly serviced. During the monitored period, the vehicle did not perform special tasks (trailer towing, motoring competitions, etc.). This is a vehicle that is not equipped with an oil quality detector, pressure gauge, or thermometer in the lubrication system. It is only equipped with an oil level sensor in the oil tank with a light signal on the driver's dashboard. As the vehicle has only basic equipment, the recording of the vehicle's operational parameters was initially solved with manual portable gauges.
This problem was gradually eliminated by the design of a custom measuring station controlled by a Raspberry Pi.4 single-board computer (Fig. 2., Fig.  4., Fig. 5.) with the possibility of data collection (motor temperature, atmospheric temperature, atmospheric humidity, distance traveled, etc.). The collected and processed data formed the basis for further research operations in R-Studio, WEKA and EXCEL Data Analysis.    The picture (Fig. 3) shows an external electronic data collection system on a Mitsubishi Lancer 1.5 Inform vehicle. This external system is not part of the CAN-BUS serial interface, nor a subsystem of another central control unit. The control unit consists of a Raspberry Pi.4.0 singleboard computer developed by the British Raspberry Pi Foundation. It operates with the Raspberry Pi OS language built on a modified Linux called Debian. The automated system is powered from the vehicle's electrical network (12 V). The system consists of a number of sensors, signal converters and regulatory elements. The display unit informs the driver "live" about the scanned physical parameters. Scanned data can be backed up to a USB memory stick or SD card. In order to connect the sensors, structural intervention was necessary, especially in the lubrication and cooling system of the vehicle.

DATA COLLECTION AND PROCESSING
The main and long-term goal of the research is the mathematical expression of the degradation processes in engine oil and the subsequent approximation of the life of the engine oil. It follows from the above that a large part of the research is focused on the collection and processing of measured data with the design of predictiveapproximation mathematical models. This process requires a voluminous statistical base of operating data from the vehicle and from the oil filling. Mathematical models are based on regularly monitored engine oil parameters (Table 1) depending on the nature of vehicle operation in the given Central European climatic conditions, which are sensed by the Raspberry Pi 4.0 measuring station (Fig. 3). The oil samples were subsequently evaluated in the tribodiagnostics laboratory at the Department of Mechanical Engineering A.O.S. Gen. M.R. Štefánik in L. Mikuláš. The tribodiagnostic analysis was focused on the basic chemical and physical properties of engine oil (Table 1.) using the most modern devices (optical and FTIR analysis) (Fig. 6).

Fig. 6. Devices for optical and FTIR analysis
The discrete Fourier transform is suitable for the analysis of stationary signals, i. e. signals that do not change in time. An oil sample is a typical stationary object of measurement where very precise measurements can be made with the help of FTIR [1].
Mitsubishi Motors Corporation recommends changing the oil after 20,000 km for these types of vehicles, and after 7,000 km in urban and extreme traffic [2].
During the research, the vehicle used Valvoline Syn Power SAE 5W-40 engine oil with ACEA A3/B3/B4 performance specification. The suitability of use is also confirmed by statistics from 2016, where up to 64% of vehicles older than 5 years were guided by this specification when choosing engine oil. In the European Union in 2016 operated approximately 126 million cars older than 5 years [3].

PROPOSAL OF APPROXIMATE MODELS FOR CALCULATING THE REMAINING LIFE OF ENGINE OIL
In this case, the proposals of mathematical models relate to the most fragile and monitored longterm parameters of TBN engine oil and AW additives (Tab. 1).
AW additives are lubricant components that chemically react with the metal surface to be protected and form a lubricating coating that protects the metal from wear under extreme lubrication conditions. Anti-wear additives are most often based on ZDDP and ashless phosphoric acid dialkyldithiophosphates, bismuth carboxylates and nano-particle potassium borates. [4] TBN is a number that characterizes the property of the oil associated with the neutralization of the acidic environment, which is created mainly during combustion and oxidation products. During operation, this ability decreases, i. e. j. the alkaline reserve decreases and the acidity of the TAN increases. The measured value should not fall below 50% of the original value of the new oil. The  decrease in this oil parameter is mainly related to the quality of the fuel (sulfur content in the diesel) and the water content in the oil. [5] [6] [7] The mathematical expression of these parameters makes it possible to create their approximate value and create predictive algorithms for the control unit of the on-board computer, which informs the driver about the current life of the engine oil. It follows from the above that, in addition to displaying data, collecting data and storing data, the designed measuring station can also be programmed with the function of approximating the life of the oil filling. However, the design of the predictor algorithm will be the goal of the next stage of research.
The essence of the calculation of the current residual life of the engine oil is based on long-term statistical measurements of the vehicle's operating conditions and regular monitoring of the chemicalphysical parameters of the oil filling, which was evaluated by laboratory devices (Fig. 6). In the search for the most optimal calculation relationship for determining the life of the oil, mutual relationships were created between the monitored parameters with the search for functional dependencies. The influence of operating factors on the complex chemical-physical picture of engine oil has been monitored and evaluated for a long time.
Since the parameters of AW additive and TBN base reserve were the weakest link in the oil, they were determined as the main indicators of engine oil life. On the basis of robust sets (collected operating data of the car and chemical-physical data of the oil filling) when applying correlation analysis and the equation of multiple regression analysis (1), the following calculation relations (2), (3) TBNn base reserve at the n th engine start [mg KOH/g] Sn n th engine start (serial number of start from the last engine oil changes) [-] ln length of the route with the vehicle at the n th engine start [km] tAn atmospheric temperature at the n th engine start [°C] tMn engine surface temperature at the n th engine start [°C] νAn relative air humidity at the n th engine start [%] In order for the approximation deviation to be as small as possible in calculations, computational applications use the method of least squares in multiple linear regression, which consists in finding the parameters of the regression function for which the sum of the squares of the deviations of the equalized values of the explanatory variable from the measured values is minimal, i. e. that it is based on minimizing the residual sum of squares. [8] The mentioned linear equation represents the notation of the n-th line of the statistical dataset. For this case, where the life of the 1st oil filling was related to 800 engine starts, the linear regression equation will be expressed in matrix form as follows: The principle of approximate calculation of the TBN parameter can be presented in matrix notation (4) for clarity. With the help of this matrix, after inserting the measured, independently transformed operating values of the vehicle, it is possible to calculate the TBN parameter of the engine oil every time the vehicle is started. In the next matrix notation, the actual measured values for the first 3 starts up to the last start no. 800, when the oil exceeded its useful life and it was replaced.
In the previous steps, a multiple linear regression was implemented, the goal of which was to obtain mathematical equations for the exact description of the degradation of the TBN parameter during the two stages of the life of the oil filling (Fig. 8, 9). The third stage of the life of the oil will be the validation stage from the research point of view. Here, the approximate-predictive function of the TBN parameter was determined based on the behavior from the previous 2nd stages and compared with the real results from the 3rd stage. The predictiveapproximation function of the TBN parameter ( Fig.  10) is created by the intersection of two equations created by regression analysis in the EXCEL Data Analysis environment from the 1st and 2nd stages.
Equation (5)   Pi) and the remaining oil life will be displayed on the driver's dashboard display (Fig. 7). Analogously in the same way as for the TBN parameter, approximation equations for the second most fragile parameter AW additives are determined using multiple regression analysis in the EXCEL Data Analysis environment. For the 1st and 2nd life stages of the oil filling for AW additives (6), (7) The third stage of the life of the oil will be the validation stage from the research point of view. Here, the approximate-predictive function of the AW additives parameter was determined based on the behavior from the previous 2nd stages and compared with the real results from the 3rd stage. The predictive -approximation function of the AW additives parameter (Fig. 11) is created by the intersection of two equations created by regression analysis in the EXCEL Data Analysis environment from the 1st and 2nd stages: Equation (8) is the second key mathematical relationship for calculating the current remaining life of engine oil. This equation will be implemented in the algorithm of the control unit (Raspberry 4.0 Pi) and the remaining oil life will be displayed on the driver's dashboard display (Fig. 7). It follows from the above that the most fragile parameters of TBN and AW additives determine the service life of the oil filling. The remaining service life of the engine oil is expressed as a percentage value on the driver's dashboard display (Fig. 7).

RESULTS AND DISCUSSION
In this case, multiple linear regression proved to be a very suitable and relatively accurate mathematical tool for calculating the degradation of the TBN parameter and AW additives in the oil. At the beginning of the research task, models were built based on TSA, but this method turned out to be inappropriate, because an important component of the time series is the seasonal component. Knowledge of the seasonal course gives us the basis for deciding how to behave in individual periods of the seasonal cycle. It is good to be prepared for high values of the analyzed variable "y" and also for its low values. [9]. The advantages of the regression method in tribological research were applied in a similar way by a team of scientists from the University of Pardubicethe benefit of the proposed and verified methodology presented in the article is a relatively simple way of determining the kinematic viscosity (i.e., 'physical' quantity) from the infrared spectrum. This spectrum can also be used not only for qualitative and quantitative analysis of chemical composition, but also for the determination of other 'non-chemical' quantities; a multivariate model must be available for them, created from a sufficient amount of quality data. The paper presents a methodology for determining kinematic viscosity 100 °C, but the process of creating a model is generally applicable to a number of other quantities (viscosity index, TBN, TAN, flash point, etc.). [10] The description of the functions of these parameters in the first two stages reached approximately 95% accuracy level. Since the oil degradation process is very complicated and difficult to predict, this result is very good. It points to correctly selected vehicle operating factors (independent variables) that enter the mathematical model. This was confirmed by correlation as well as regression analysis. It is good that it was implemented on the parameters that are the weakest links of the oil.
As for the prediction itself in the third validation stage, an accuracy in the range of 80 to 90% is an expected and satisfactory result. The fact is that the data measured for prediction were based on only two stages. It is assumed that if this research were to be continued (data would be collected from 3 to 4 stages), the accuracy of the prediction could reach values of 90% or more based on this methodology. Based on regression analysis, the number of engine starts and engine temperature were shown to be the main factors of dependence in relation to oil degradation. To verify the accuracy, the relative measurement error x method was used:  The relative measurement deviation x is determined as the ratio of the absolute value of the absolute measurement deviation and the conventional true value of the measured quantity. It represents a positive numerical value, often expressed as a percentage [11]. xv calculation result (value of the calculated value according to the equation) xm measured value (real measured value) When balancing the results, it was additionally shown that, with some similarity, General Motors Corporation-GM Oil Life System is also developing predictive systems for its cars, which also has algorithms with systems of linear equations built into the control units of the vehicles. The Daimler-Chrysler Corporation Flexible System (FSS) or the oil quality evaluation system from the Ford Motor Company also show a certain, though small, similarity in the equations.
However, their accuracy remains questionable. These electronic systems indirectly monitor oil life based on engine operating conditions, i.e. j. they do not measure oil quality by contact method. This is due to the fact that it is difficult for contact sensors of the oil condition to monitor the complex chemical picture of the oil in real time (on-line) when the vehicle is running. [12] [-] When comparing our mathematical model (10), (11) with a similar system General Motors Corporation GM Oil Life System (12), it can be concluded that our system is focused primarily on the number of engine starts and the climatic environment of the vehicle being operated. The GM Oil Life System is a significantly more complex engine oil life calculation system, where data collection is mainly focused on the engine and its accessories during operation.
Most predictive models are built on kilometer intervals, ie sampling and analysis of the sample is carried out exactly after the observed kilometer interval (1.000 km, 2.000 km, etc.), regardless of the time duration of the interval. The difference between our system and the others lies, among other things, in the fact that the sensing of parameters is exclusively before engine start (many systems read operating variables continuously during engine load), as cold vehicle starts have the greatest impact on oil degradation [13].
It cannot be unequivocally stated that oil quality detection systems have successfully proven themselves in practice. Usually, if the systems are